Main Content

Stress Tests: Restore Compliance with the APA

Supervisory practices at the Federal Reserve must be reformed to restore compliance with the Administrative Procedure Act, especially for stress tests.

By Hal Scott

Since the financial crisis ended, the authority of banking regulators, particularly the Federal Reserve, has greatly expanded to include the power to disapprove of distributions to shareholders under the Fed’s annual stress tests and to break up large banks if their living wills are deemed to be not credible. The proper exercise of such powers requires that the Fed adopt and implement rules that govern these decisions. (Note: This article, which is derived from a report published by the Committee on Capital Markets Regulation entitled “The Administrative Procedure Act and Federal Reserve Stress Tests,” primarily focuses on the Fed’s stress tests’ compliance with federal administrative law.)

In general, when federal agencies adopt rules, the Administrative Procedure Act of 1946 (APA) requires that agencies: (1) provide the public notice of proposed rules and an opportunity to comment on them (“notice-and-comment procedures”); (2) publish their final rules (Id. § 552(a)(1)(D)); and (3) have their actions be subject to judicial review and reversal if they fail to comply with APA procedural requirements or are otherwise arbitrary and capricious (Id. § 706(2)).

Fed compliance with these procedural requirements is of great importance in the context of the annual quantitative stress tests it conducts as part of its Comprehensive Capital Analysis and Review (CCAR). The stress tests quantify whether a bank can maintain sufficient capital to meet regulatory minimums under hypothetical future adverse conditions. Importantly, as the Treasury stated in its June 2017 report to the president on financial reform, the stress tests act as a de facto binding capital constraint, particularly for large banks.

The Fed has not complied with the APA’s procedural requirements in adopting key aspects of the stress tests, even though doing so would reduce the threat of legal challenge to the Fed’s actions.

That’s because if a bank fails to maintain adequate capital under the stress tests’ projections, then the Fed can prevent the bank from returning cash to shareholders through dividends and stock repurchases, and the amount of capital necessary to survive these tests will be greater than that necessary to comply with the U.S. Basel capital requirements.

In my view, the Fed has not complied with the APA’s procedural requirements in adopting key aspects of the stress tests, even though doing so would reduce the threat of legal challenge to the Fed’s actions and result in better public policy outcomes. In this article, I will (1) describe the Fed’s stress test process; (2) explain how that process likely fails to comply with APA procedural requirements; and (3) make recommendations on how to improve the process to restore compliance with the APA, including that key components of the stress tests be subject to public notice and comment.

I. The Fed’s Stress Tests

A. Overview
The Fed runs its own quantitative stress tests of bank holding companies with total consolidated assets equal to or greater than $50 billion and non-financial companies supervised by the Fed. The stress tests serve “as a means for assuring that large, complex financial institutions have sufficient capital to allow them to remain viable intermediaries even under highly stressful conditions,” according to Governor Daniel K. Tarullo, Members of the Board of Governors of the Federal Reserve System (Stress Testing After Five Years. June 25, 2014). Banks subject to the tests hold approximately 75% of U.S. banking assets. Adoption of the Treasury Department’s recommendation that the threshold be tailored to apply only to certain banks based on factors such as business models and organizational complexity would reduce the burden the tests impose, particularly on smaller banks that are primarily engaged in commercial banking activities (U.S. Department of the Treasury, supra note 5, at 12).

The stress test process occurs over a full year. The Fed gathers historical bank-specific information throughout the year, discloses the hypothetical adverse economic scenarios it will apply in the tests by mid-February (12 CFR § 225.42(b)), and collects the banks’ capital plans describing each bank’s planned stock and debt repurchases and issuance, as well as dividend payments, in early April (Id. § 225.8(e)). Using this information, including detailed bank-specific data on loans, investments, revenues, operational risks, trading activities, net interest income, and assets, the Fed runs its own internal models to project a bank’s regulatory capital ratios under the hypothetical adverse economic conditions and releases the results by June 30 each year (Id. § 225.8(f)(2)(v)).

The Fed compares the projected capital ratios of each bank to the regulatory minimum capital ratios based on Basel standards and will object to a bank’s planned dividends and stock repurchases if the projected capital ratio falls below the regulatory minimum capital ratio in any of the four calendar quarters immediately following the release of the results (CCAR 2016).

There are three key aspects of the tests: (1) assumptions on economic and financial market conditions; (2) models projecting incomes (which must predict losses); and (3) minimum capital ratios.

B. Stress Test Assumptions
The Fed must decide what stressful economic and financial market conditions banks might encounter and so must develop assumptions about these conditions. These assumptions include estimates of future gross domestic product (GDP), interest rates, unemployment rates, and dozens of other macroeconomic and financial market variables (Hirtle, Beverly, and Andreas Lehnert. Supervisory Stress Tests, Federal Reserve Bank of New York Staff Reports, No. 696 at 18. 2014). The assumptions are applied consistently among the banks subject to the stress tests. For example, in the 2017 tests, the Fed’s worst-case scenario assumed the growth rate of GDP would decline 10.6 percentage points from the fourth quarter of 2016 to the second quarter of 2017 and the unemployment rate would hit 10% by mid-2017.17 Those are extreme assumptions compared to the current GDP growth rate of 2% and an unemployment rate of less than 5%.

The Fed develops three sets of assumptions that are referred to as “baseline,” “adverse,” and “severely adverse” scenarios. The severely adverse assumptions are the harshest and are therefore the most important because banks must meet the minimum regulatory capital ratios in projections made under all three sets of assumptions. The Treasury’s proposal to eliminate the adverse scenario makes sense because it has no real consequences.

Despite the importance of the economic scenarios to the stress tests, they have not been subject to a notice-and-comment rulemaking procedure. And although the Fed has published a scenario design paper outlining its general approach to developing the assumed economic and financial market conditions, it does not provide details about each year’s assumptions as they are being developed and does not seek public comment on the assumptions (See generally, Policy Statement on the Scenario Design Framework for Stress Testing. 78 Fed. Reg. 71,435. Nov. 29, 2013). Instead, the assumptions are disclosed to the public as a fait accompli and are required to be applied to the banks.

C. Stress Test Models
With set assumptions and bank-specific data in hand, the Fed determines how each bank’s capital regulatory ratios would fare. To do so, the Fed applies its own internally developed projection models. It does so in a secretive and opaque manner.

The Fed’s models fall into five broad categories: (1) models to project losses on loans held in the accrual loan portfolio; (2) models to project other types of losses, including those from changes in the fair value of loans held for sale, losses on securities, trading, and counterparty exposure, and losses related to operational risk events; (3) models to project the components of pre-provision net revenue; (4) models to project balance sheet items and risk-weighted assets; and (5) calculations to project capital ratios given projections of pre-tax net income and provisions for the allowance for loan and lease losses.

Despite the importance of the economic scenarios to the stress tests, they are not set forth in rules that are subject to notice-and-comment procedures.

The Fed has never released the models to the public either, before the stress tests are conducted or after the results are released. While the Fed makes some effort to provide transparency by including a high-level summary of its modeling approach in the release of the stress test results, such disclosures do not provide specific details about the models. And although a Model Validation Council currently made up of six outside academics evaluates the models each year, the Fed does not disclose the feedback it receives from that body.

Fed officials have justified the secrecy surrounding the models by contending that disclosure of the models would allow banks to “game” the tests – presumably by, among other things, accumulating assets after disclosure of the models that perform well on the test and then changing the portfolio composition after the test is over.

D. Capital Ratios
The final step in the stress test is for the Fed to compare a bank’s projected capital ratios with the required regulatory minimums. Unlike the assumptions and models, the applicable capital ratios are set forth in Fed rules that have been subject to notice-and-comment procedures. The applicable ratios are the tier 1 leverage ratio, the supplementary leverage ratio (added in the 2017 stress tests), the common equity tier 1 ratio, the tier 1 risk-based capital ratio, and the total risk-based capital ratio. On at least six occasions, the Fed has objected to a bank’s capital plan because at least one of its projected capital ratios fell below the regulatory minimum. Currently, the Fed is considering adding the global systemically important bank (G-SIB) surcharge to the capital required of the eight G-SIB banks.

II. The Fed’s Stress Tests and Compliance with the APA

The Fed-developed assumptions and models are central to the stress-testing process, but they’re not subject to input from the public or the firms affected by them. That failure to seek public input likely violates the notice-and-comment procedural requirements of the APA because the assumptions and models constitute rulemakings that are not subject to a valid exemption.

A. The Fed’s Assumptions and Models Are Rulemakings
The APA divides agency actions into two categories: rulemakings and adjudications. Critically, rules are generally subject to the APA’s notice-and-comment procedures and adjudications are not.

The key distinction between rules and adjudications is that rules relate to policy considerations and have future effect, whereas adjudications are backward looking and usually resolve disputed facts in a specific case (U.S. Department of Justice. Attorney General’s Manual of the Administrative Procedure Act. 14-15. 1947 (stating that rulemaking is “agency action which regulates future conduct” and “is primarily concerned with policy considerations”); United States v. Fla. E. Coast Rwy. Co., 410 U.S. 224, 245 (1973) (stating that rulemakings are “proceedings for the purpose of promulgating policy-type rules or standards”). The assumptions and models are best categorized as rules for at least two reasons. First, the assumptions about economic conditions and models on the predicted responses to those conditions are applied to all U.S. banks with assets over $50 billion and therefore have policy implications as they establish the de facto minimum capital requirement for 75% of the U.S. banking system. Second, the assumptions and models have future effect through their significant influence over whether banks will be allowed to make future capital distributions.

If regulators are using pre-existing criteria to judge the credibility of living wills that have not been made public and subject to notice-and-comment procedures, then those criteria should be treated as rules and subject to notice-and-comment procedures.

The Fed, however, could argue that the assumptions and models are part of an adjudication – a bank-by-bank assessment of whether an individual bank is well capitalized enough to make planned capital distributions. It could point to the Supreme Court’s 1974 case NLRB v. Bell Aerospace, in which the Court held that, in an adjudication about whether to certify a union, the NLRB could determine who was a “managerial employee,” a question central to the certification determination, even if the definition would subsequently apply more broadly in the future to other employers.

However, that case is easily distinguishable. In Bell Aerospace, the NLRB had not pre-defined the term “managerial employee” and had to apply a definition to reach a judgment in the case. In the Fed’s review of a bank’s capital adequacy, however, the assumptions and models are developed in advance of and separately from the conclusion about the capital adequacy of an individual bank. The assumptions and models are criteria that function as rules that the Fed has already developed and that it consistently applies across banks. Indeed, the assumptions and models function no less as rules than the capital ratio requirements, which the Fed treats as rules.

B. Exceptions to the APA’s Notice-and-Comment Procedures Do Not Apply
If the assumptions and models are rulemakings, then the APA notice-and-comment procedures govern unless one of three relevant exceptions apply: (1) they are merely guidance; (2) they are “interpretative” rules; or (3) the Fed demonstrated that the notice-and-comment procedures are unnecessary, impractical, or contrary to the public interest (the “good cause exception”).

1. The Exceptions Do Not Apply to the Assumptions

First, the guidance exception does not apply to the assumptions because they are treated by the Fed as binding. The D.C. Circuit has held that guidance cannot be treated by the agency as binding. The Fed applies the assumptions uniformly across all of the banks and provides no process whereby a bank can discuss with the Fed why a particular assumption should not be applied to it or why a different assumption might be more relevant.


Second, the assumptions do not meet the D.C. Circuit’s definition of an “interpretative” rule, which is one that “merely interprets” and “does not itself purport to impose new obligations or requirements on regulated parties.” The Fed’s assumptions are not based on an interpretation of an existing rulemaking and the Fed’s assumptions impose new obligations on regulated parties because they significantly affect the amount of capital a bank needs to hold in a given year.


Finally, the good cause exception is narrowly construed by courts and inapplicable. The D.C. Circuit has explained that the three conditions for the good cause exception are as follows: It is (1) impractical for an agency to follow notice-and-comment procedures when an agency finds that due and timely execution of its functions would be impeded by following the procedures; (2) unnecessary to follow procedures if the rule is routine in determination, insignificant in nature and impact, and inconsequential to the industry and public; and (3) contrary to the public interest if the public interest would be defeated by compliance with the procedural requirements.


There is no apparent reason why the good cause exception would apply. The Fed can develop proposed assumptions with sufficient time for the Fed to incorporate public comments prior to issuing the final assumptions. Moreover, for reasons already discussed, the assumptions are not insignificant or inconsequential. Finally, public participation through notice-and-comment procedures would be in the public interest because it would allow the public to provide informational feedback to the Fed, thus enhancing the quality of the assumptions.


2. The Exceptions Do Not Appear to Apply to the Models

In regard to the models, the exceptions for guidance and interpretative rules do not apply because the models are binding, do not simply interpret an existing rulemaking, and impose substantive new obligations to hold capital based on the design of the models themselves. However, there is possibly some basis to conclude that the good cause exception may apply.


There is a dearth of case law directly on this point, but two 40-year-old cases from a now-defunct court of appeals held that notice-and-comment procedures could be bypassed where pre-announcing changes to the price of regulated oil prices would have undermined the aims of the rules.42 Likewise, if the stress test models can be “gamed” then disclosing them prior to the tests could arguably undermine the Fed’s ability to accurately predict how banks would truly be affected by stressful conditions.


That argument has serious problems. First, it is unclear how big of a concern gaming really is. For example, it would be highly costly and impractical for a bank to significantly change risk exposures in its banking book in a matter of months. Second, the Fed could address gaming by adopting and enforcing a good faith or anti-evasion principle.


The Fed also has to contend with the separate requirement that final rules be published once finalized (a requirement the Fed does not satisfy with respect to the models). The good cause exception for notice-and-comment procedures is distinct from the requirement that final rules be published, so even if the good cause exception applied, the models would need to be published when they’re final.


III. Enhancing Compliance with the APA


The Fed should comply with the APA’s notice-and-comment procedures in adopting the economic and financial market assumptions used in the annual stress tests and in the models it uses to make its projections.


Doing so would allow the public to provide valuable information and insight, which will help promote quality decision making when it comes to determining the appropriate assumptions.43 Additionally, it would alleviate the concern raised by the Treasury that “[t]he lack of transparency ... can make it difficult for firms to efficiently allocate capital across products and exposures.” Finally, it would reduce the threat of legal challenge to Fed actions taken in reliance on the stress tests, including the rejection of a bank’s capital plan.


It is crucially important that development of both the assumptions and models involve public input because any substantive changes made to the assumptions could be offset by changes to the secret models. The Treasury has proposed conducting the tests once every two years, and doing so could give the Fed adequate time to comply with the APA.


Moreover, although I have focused on the quantitative aspect of the stress test in this article, the Fed also qualitatively reviews banks’ capital planning abilities and can object to shareholder distributions on qualitative grounds. That review process is also very opaque, and I agree with the Treasury Department that such a review should be conducted as part of the normal supervisory process rather than the review and approval of capital distributions.


More generally, the Fed should be diligent to ensure that it is complying with the APA when engaged in other supervisory activities, such as supervising the “living wills” process. If regulators are using pre-existing criteria to judge the credibility of living wills that have not been made public and subject to notice-and-comment procedures, as the GAO has found, then those criteria should be treated as rules and subject to notice-and-comment procedures.


Complying with the APA is not only the law, it is good public policy. It promotes transparency and accountability, increases the quality of agency decision making, and reduces uncertainty about the lawfulness of agency actions.


About The Author:

Hal S. Scott is the Nomura Professor and Director of the Program on International Financial Systems (PIFS) at Harvard Law School, where he has taught since 1975.


Scott has a B.A. from Princeton University (Woodrow Wilson School, 1965), an M.A. from Stanford University in political science (1967), and a J.D. from the University of Chicago Law School (1972). In 1974–1975, before joining Harvard, he clerked for Justice Byron White.


Scott is the Director of the Committee on Capital Markets Regulation, a bipartisan nonprofit organization dedicated to enhancing the competitiveness of U.S. capital markets and ensuring the stability of the U.S. financial system via research and advocacy. He is also a member of the Bretton Woods Committee, a member of the Market Monitoring Group of the Institute of International Finance, a past Independent Director of Lazard, Ltd. (2006–2016), a past President of the International Academy of Consumer and Commercial Law, and a past Governor of the American Stock Exchange (2002–2005).